\documentclass[12pt]{article}
\usepackage[utf8]{inputenc}
\usepackage{setspace} % to make double space
\usepackage{footmisc}
\setlength{\footnotesep}{\baselineskip} %use 1.67\baselineskip for a double space
\makeatother
\renewcommand{\footnotelayout}{\doublespacing}

%\renewcommand{\floatpagefraction}{.8}%prevent figure only pages

\usepackage[margin= 1 in]{geometry}
  \usepackage{placeins} %Forcing table/figure placement within sections and subsections

%\usepackage[round]{natbib} % Changes text citation from brackets to parentheses
\usepackage{apacite}
\usepackage{graphicx} % To include images from pdf and png
\usepackage{amssymb, amsmath} % Math expressions
\usepackage{bbding} %checkmarks
\usepackage{float}
\usepackage{wrapfig}
\usepackage{authblk} % To organize author info in title page
\usepackage{rotating}
\usepackage[labelfont=bf]{caption}
\usepackage{multirow}
\usepackage{longtable}
\usepackage{tocloft}

\addtolength{\cftsecnumwidth}{8pt}

\newcommand{\todo}[1]{\vspace{5 mm}\par \noindent
\marginpar{\textit{To Do}}
\framebox{\begin{minipage}[c]{0.95 \textwidth}
\tt\begin{center} #1 \end{center}\end{minipage}}\vspace{5 mm}\par}




\usepackage{hyperref}
\hypersetup{
    colorlinks,
    citecolor=black,
    filecolor=black,
    linkcolor=blue,
    urlcolor=black
}

\allowdisplaybreaks 
\raggedbottom


\begin{document}

\pagenumbering{arabic}% resets `page` counter to 1
\renewcommand*{\thepage}{A\arabic{page}}
\appendix
\renewcommand{\thesection}{A\arabic{section}}  
\renewcommand{\thetable}{A\arabic{table}}  
\renewcommand{\thefigure}{A\arabic{figure}}


\singlespacing
\begin{center}
\section*{Appendix to the Paper: \\
\medskip
 Elite Change without Regime Change: Authoritarian Persistence in Africa and the End of the Cold War}
\end{center}

\thispagestyle{empty}
\setcounter{page}{0}
\tableofcontents
\clearpage

\FloatBarrier

%\section{Variables}

%\begin{table}[!hbt]
%\begin{tabular}{|p{3cm}|p{13cm}|}
%\hline
%\textbf{Variable} & \textbf{Description} \\
%\hline
%CWE\_5 & Cold War end period variable coded 1 for years 1988-1992.  \\
%\hline
%CWE\_4 & Cold War end period variable selecting the four years with the most cabinet turnover within the 1988-1992 window.  \\
%\hline
%CWE\_3 & Cold War end period variable selecting the three years with the most cabinet turnover within the 1988-1992 window.   \\
%\hline
%CWE\_2 & Cold War end period variable selecting the two years with the most cabinet turnover within the 1988-1992 window.   \\
%\hline
%CWE\_1 & Cold War end period variable selecting the year with the most cabinet turnover within the 1988-1992 window.   \\
%\hline
%ln\_oilgas\_rents & The natural log of oil and gas rents from \cite{Ross}.\\
%\hline
%pers2\_pl &The continuous measure of personalism from  \cite{Wright}. \\
%\hline
%lcgdppc & The natural log of GDP per capita measured by \cite{maddison}.\\
%\hline
%cabinet\_size & The number of individuals in the cabinet.  Individuals holding multiple cabinet portfolios are only counted once. Leaders are excluded from cabinet size. \\
%\hline
%left\_truncated & WhoGov begins collecting cabinet data in 1966. This means that tenure of ministers present in 1966 in countries that were independent prior to 1966 is unknown (left truncated).  This variable is coded 1 for ministers whose tenure is left truncated and 0 for all other ministers. \\
%\hline
%minister\_tenure & The number of consecutive years an individual has held a cabinet level position.  This variable is not portfolio specific. Therefore, a minister who enters the cabinet as Minister of Defense in year $t$ and is reassigned as Minister of Interior in year $t+1$ has a tenure of 2 years in year $t+1$.  Minister tenure is calculated using data from WhoGov \cite{Nyrup}. \\
%\hline
%leader\_tenure & The number of years the state leader has been in office.  Leader data comes from WhoGov \cite{Nyrup} and is cross-checked with Archigos \cite{archigos}.\\
%\hline
%leader\_exit & Leader exit is coded 1 in the final year the leader appears in the dataset and 0 otherwise. \\
%\hline
%election\_corrected & This variable is coded 1 for each year in which a legislative or executive election takes place and 0 otherwise using data from NELDA \cite{hydemarinov}. However, the structure of the WhoGov data require corrections of election years in certain cases. See Section \ref{corrections} for details on corrections to election years. \\
%\hline
%coup\_corrected & This variable is coded 1 for each year in which at least one failed coup attempt took place and 0 where no failed coup attempts took place. Data are take from \cite{powellthyne}. As with elections, the year in which a failed coup attempt takes place must be corrected in some instances because of the coding of the WhoGov data. See Section \ref{corrections} for details. \\
%\hline

%\end{tabular}
%\end{table}

%\FloatBarrier

\section{Coding Corrections to GWF Regime Type,  Elections, and Failed Coups}\label{corrections}

WhoGov's collection of cabinet data at the beginning of July each year complicates the coding of regime type and the Election and Failed coup variables.  Each of these variables cannot be merged on a country-year basis without errors.  Below, we explain the coding corrections made for regime type, elections, and failed coups.

We use Geddes, Wright, and Frantz's (2014) (hereafter, GWF) data to identify our sample of autocracies in sub-Saharan Africa. However, the country-year coding procedure for regimes used by GWF is not perfectly compatible with the cabinet observations each July from WhoGov.  When merging these datasets by country-year,  errors sometimes occur in years where one regime exits and another begins.  This is because the country-year version of GWF records regimes as beginning on January 1 after the actual regime change date.  For example, a regime that begins on August 1, 2000 would not show up in the GWF country-year data until 2001.  In that example,  a country-year merge with WhoGov would be correct. However, if the true regime start date was March 1, 2000, the country-year merge would fail in 2000. The WhoGov cabinet observed in July 2000 would be merged with the GWF regime from 2000 which would have already exited power in July.  Since GWF record the true start and end dates of regimes, we are able to correct these problematic cases, ensuring that the cabinet observations are matched up with the appropriate regime. 

A similar issue occurs with the coding of elections and failed coups. We recode the timing of elections and failed coup attempts that occur in year $t$ but prior to July as occurring in year $t-1$. This coding reflects the assumption that cabinet reshuffles can only be attributed to elections and failed coup attempts if they take place directly following these events. Consider the following example:

\begin{quote}
    Minister X is present in the cabinet in 2005, but not in 2006. Under our coding scheme, Minister X is coded as exiting the cabinet in 2005. Now, assume that Minister X's country held an election in February of 2005. If Minister X was dismissed following the election in February of 2005, they would not have been recorded as a cabinet member in 2005 since the observation is taken at the beginning of July. As such, we code the election as taking place in 2004. This prevents the dismissal of Minister X in 2005 from being attributed to the election. If, however, the 2005 election had taken place in September of 2005, the election year would remain coded as 2005.
\end{quote}

\FloatBarrier

\clearpage

\section{Descriptive Statistics}\label{sec:descriptives}

\input{samples_snapshot.tex}
\FloatBarrier
\subsection{Geddes, Wright, and Frantz (GWF) Autocracies Samples}

\input{gwf_ssa_stats.tex}
\FloatBarrier
\input{gwf_ssa_sample.tex}
\FloatBarrier

\input{gwf_all_stats.tex}
\FloatBarrier

\input{gwf_all_sample.tex}
\FloatBarrier

\subsection{Lexical Index of Electoral Democracy (LIED) Autocracies Samples}

\input{lied_ssa_stats.tex}
\FloatBarrier

\input{lied_ssa_sample.tex}
\FloatBarrier

\input{lied_all_stats.tex}
\FloatBarrier

\input{lied_all_sample.tex}
\FloatBarrier

\subsection{Varieties of Democracy Regimes of the World (VDem ROW) Autocracies Samples}
\input{row_ssa_stats.tex}
\FloatBarrier

\input{row_ssa_sample.tex}
\FloatBarrier

\input{row_all_stats.tex}
\FloatBarrier

\input{row_all_sample.tex}
\FloatBarrier

\subsection{GWF Balanced Sample}

The balanced sample selects countries with periods of time continuously defined as autocracies by Geddes, Wright, and Frantz (2014) up to the end of the Cold War and retains those countries in the sample until 2010, even if they transition to democracy. \footnote{WhoGov is missing cabinet data for Malawi in 1983 and Rwanda in 1973.}

\input{gwf_ssa_balanced_sample.tex}
\FloatBarrier

\input{gwf_ssa_balanced_stats.tex}
\FloatBarrier

\subsection{CWE Surviving Leaders Sample }

Defining leaders as surviving the Cold War end  (CWE) period is not straightforward. As such, we adopt a conservative approach. We define the sample of CWE surviving leaders as leaders with at least three years of tenure before the beginning of our CWE period (1988) and at least three years of tenure after our CWE period (1992). \\

\input{cwe_survivors.tex}
\FloatBarrier

\clearpage

\section{Models of Minister Survival Across Samples and Specifications}

\subsection{Pooled Logit Models}

The average marginal effects plotted in Figure 3 of the main text are calculated using Model 1 in Tables \ref{tab:cwe5_pooled_none}, \ref{tab:cwe5_pooled_baseline}, \ref{tab:cwe5_pooled_tenure}, and \ref{tab:cwe5_pooled_events}.  We create Figure 4 in the main text  using Model 1 in Table \ref{tab:cwe5_pooled_events}. In each table, Models 2-6 estimate the same specification in Model 1, but using samples with alternative definitions of autocracy and different subsets of countries.  The column labels indicate the sample used. Descriptive statistics for each sample are in Section \ref{sec:descriptives}


\input{cwe5_pooled_none.tex}


\input{cwe5_pooled_baseline.tex}

\input{cwe5_pooled_tenure.tex}

\input{cwe5_pooled_events.tex}


{\renewcommand\normalsize{\small}%
\normalsize\input{cwe5_pooled_extras}}

\FloatBarrier

\subsection{Conditional Logit Models (Leader Fixed Effects)}

\input{cwe5_fe_none.tex}

\input{cwe5_fe_baseline.tex}

\input{cwe5_fe_tenure.tex}

\input{cwe5_fe_events.tex}

\input{cwe5_fe_extras.tex}


\FloatBarrier

\subsection{Models of Minister Survival with Balanced Sample}

\input{cwe5_pooled_balanced.tex}

\FloatBarrier

\input{cwe5_fe_balanced.tex}

\FloatBarrier


\subsection{Models of Minister Survival with CWE Surviving Leaders Sample}
\FloatBarrier

\input{survivors_pooled.tex}

\FloatBarrier


\input{survivors_fe.tex}

\FloatBarrier


\section{Senior Minister Exit Analyses}

\subsection{Pooled Logit Models from Figure 5}

\input{cwe_senior_pooled.tex}
\FloatBarrier

\subsection{Pooled Logit Models of Minister Survival, Senior Minister Only Sample}

\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{Fig2_Senior_AMEs.pdf}
    \caption{\textbf{Average Marginal Effects of CWE with 95\% Confidence Intervals Limiting the GWF Sub-Saharan Sample to Senior Ministers}}
    \label{fig:leader_minister_exits}
\end{figure}

\input{cwe5_senior_pooled.tex}

\input{cwe4_senior_pooled.tex}

\input{cwe3_senior_pooled.tex}

\input{cwe2_senior_pooled.tex}

\input{cwe1_senior_pooled.tex}
\FloatBarrier

\subsection{Conditional Logit (Leader Fixed Effects) Models of Minister Survival, Senior Minister Only Sample}

\input{cwe5_senior_fe.tex}

\input{cwe4_senior_fe.tex}

\input{cwe3_senior_fe.tex}

\input{cwe2_senior_fe.tex}

\input{cwe1_senior_fe.tex}
\FloatBarrier



\section{Coercive Portfolio Minister Exit Analyses}

\subsection{Pooled Logit Models from Figure 6}

\input{cwe_coercive_pooled.tex}
\FloatBarrier

\subsection{Pooled Logit Models of Minister Survival,  Coercive Portfolio Minister Only Sample}

\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{Fig2_Coercive_Portfolio_AMEs.pdf}
    \caption{\textbf{Average Marginal Effects of CWE with 95\% Confidence Intervals Limiting the GWF Sub-Saharan Sample to Coercive Portfolio Ministers}}
   %\label{fig:leader_minister_exits}
\end{figure}

\input{cwe5_coercive_pooled.tex}

\input{cwe4_coercive_pooled.tex}

\input{cwe3_coercive_pooled.tex}

\input{cwe2_coercive_pooled.tex}

\input{cwe1_coercive_pooled.tex}
\FloatBarrier

\subsection{Conditional Logit (Leader Fixed Effects) Models of Minister Survival, Coercive Portfolio Minister Only Sample}

\input{cwe5_coercive_fe.tex}

\input{cwe4_coercive_fe.tex}

\input{cwe3_coercive_fe.tex}

\input{cwe2_coercive_fe.tex}

\input{cwe1_coercive_fe.tex}
\FloatBarrier



\section{High Prestige Minister Exit Analyses}
\FloatBarrier
\subsection{Pooled Logit Models}

%We examine the fates of high prestige versus medium and low prestige ministers in the same manner that we compared junior and senior ministers and coercive portfolio and other portfolio ministers in the main text of our paper. We estimate models of minister exit for each CWE window size,  interacting CWE with the high prestige dummy.  We then calculate the unconditional first difference in the probability of minister exit for the CWE and non-CWE periods for ministers in both high prestige and medium/low prestige portfolios. The first difference estimates presented in Figure \ref{fig:lowmed_high_diff} are very similar to those in Figure 5 of the main text. Using our least favorable 5-year CWE window, ministers in low and medium prestige portfolios experienced an estimated 8.3 percentage point increase in exit probability and those in high prestige portfolios a 8.5 percentage point increase. The estimated first differences for both high prestige and medium/low prestige ministers increase as the CWE coding becomes more favorable, and the estimates for high and medium/low prestige ministers at each CWE window size are very similar one another. Under our most favorable CWE window, low/medium prestige ministers face an estimated 39 percentage point increase at the end of the Cold War and high prestige ministers a 35 percentage point increase. This provides additional evidence that the changes in elite coalitions documented in Figure 3 of the main text were not limited to those in the least important or influential positions.



%first difference plot from old main text figure 6
\begin{figure}[!hbt]
   \centering
   \includegraphics[width = 1\textwidth]{LowMed_High_Diff.pdf}
    \caption{\textbf{Difference in minister exit probabilities at CWE = 1 and CWE = 0 for ministers in high prestige and medium/low prestige portfolios with 95\% confidence intervals.  First differences calculated using models from Table \ref{tab:cwe_high_pooled}. }}
   \label{fig:lowmed_high_diff}
   \end{figure}
   
\input{cwe_high_pooled.tex}
\FloatBarrier



\subsection{Pooled Logit Models of Minister Survival, High Prestige Minister Only Sample}

\begin{figure}[!hbt]
   \centering
   \includegraphics[width = 1\textwidth]{Fig2_High_Prestige_AMEs.pdf}
          \caption{\textbf{Average Marginal Effect of CWE Window Variables with 95\% Confidence Intervals,  High Prestige Minister Only Sample}}
   \label{fig:Fig2_High_Prestige_AMEs}
   \end{figure}
\FloatBarrier

{\renewcommand\normalsize{\small}%
\normalsize\input{cwe5_high_pooled.tex}}

{\renewcommand\normalsize{\small}%
\normalsize\input{cwe4_high_pooled.tex}

{\renewcommand\normalsize{\small}%
\normalsize\input{cwe3_high_pooled.tex}

{\renewcommand\normalsize{\small}%
\normalsize\input{cwe2_high_pooled.tex}

{\renewcommand\normalsize{\small}%
\normalsize\input{cwe1_high_pooled.tex}

\FloatBarrier


\subsection{Conditional Logit (Leader Fixed Effects) Model of Minister Survival, High Prestige Minister Only Sample}

\input{cwe5_high_fe.tex}

\input{cwe4_high_fe.tex}

\input{cwe3_high_fe.tex}

\input{cwe2_high_fe.tex}

\input{cwe1_high_fe.tex}

\FloatBarrier




\section{Yearly Leader and Minister Turnover}
\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{leader_minister_exits.pdf}
    \caption{\textbf{Leader and Minister Exits across GWF Sub-Saharan Sample}}
    \label{fig:leader_minister_exits}
\end{figure}

\FloatBarrier


\section{Additional Placebo Tests}

\subsection{Five Year Placebo Test,  Conditional Logit (Leader Fixed Effects)}
\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{Fig5_cwe5_placebo_fe.pdf}
    \caption{\textbf{Five-Year Window Placebo Test: CWE Coefficients with 95\% Confidence Intervals from Conditional Fixed Effects Logit Model (Leader Fixed Effects)}}
    \label{fig:Fig5_cwe5_placebo_fe}
\end{figure}
\FloatBarrier

\clearpage
\subsection{Three Year Placebo Test,  Pooled Logit}
\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{Fig5_cwe3_placebo.pdf}
    \caption{\textbf{Three-Year Window Placebo Test: CWE Coefficients with 95\% Confidence Intervals}}
    \label{fig:Fig5_cwe3_placebo}
\end{figure}

\FloatBarrier
\clearpage
\subsection{Three Year Placebo Test,  Conditional Logit (Leader Fixed Effects)}
\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth]{Fig5_cwe3_placebo_fe.pdf}
    \caption{\textbf{Three-Year Window Placebo Test: CWE Coefficients with 95\% Confidence Intervals from Conditional Fixed Effects Logit Model (Leader Fixed Effects)}}
    \label{fig:Fig5_cwe3_placebo_fe}
\end{figure}
\FloatBarrier

\clearpage


\section{Logit Model of Senior Minister Exit Across Cold War,  Cold War End, and Post-Cold War}

\input{senior_cwe_pcw.tex}

\FloatBarrier

%\section{Additional Analyses of Minster Exit Across Minister Tenure}
%GRAPHS FROM ORIGINAL SUBMISSION PLUS TABLES OF RESULTS

%\begin{figure}[!hbt]
%    \centering
%   \includegraphics[width = 1\textwidth]{cwe5_int_exit_probs.pdf}
 %   \caption{\textbf{Predicted Probability of Minister Exit with 95\% Confidence Intervals,  MODEL AND TABLE NUMBER HERE}}
 %   \label{fig:cwe5_int_probs}
%\end{figure}
%\FloatBarrier



%\begin{figure}[!hbt]
%    \centering
%   \includegraphics[width = 1\textwidth]{cwe1_int_exit_probs.pdf}
 %   \caption{\textbf{Predicted Probability of Minister Exit with 95\% Confidence Intervals,  MODEL AND TABLE NUMBER HERE}}
 %   \label{fig:cwe1_int_probs}
%\end{figure}
%\FloatBarrier


\section{Models Interacting CWE and Oil/Gas Rents}

\input{cwe5_oilgas_pooled.tex}

\input{cwe5_oilgas_fe.tex}

\FloatBarrier

\section{Models of Included Ethnic Groups,  Included Parties, and Cabinet Size}

\input{other_dvs.tex}

\FloatBarrier

\section{Cabinet Plots by Country}

The plots below replicate Figure 1 of the main text for all countries and leaders in the GWF SSA dataset.  These plots differ only by the inclusion of vertical dashed lines for leader exits. 

\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 1]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 2]{All_Triangles.pdf}
\end{figure}
\FloatBarrier



\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 3]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 4]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 5]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 6]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 7]{All_Triangles.pdf}
\end{figure}
\FloatBarrier

\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 8]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 9]{All_Triangles.pdf}
\end{figure}
\FloatBarrier

\begin{figure}[!hbt]
    \centering
   \includegraphics[width = 1\textwidth, page = 10]{All_Triangles.pdf}
\end{figure}
\FloatBarrier


\bibliography{References}

\end{document}